<![CDATA[This research focuses on the application of computer vision technology in smart agriculture, particularly for detecting the ripeness level of coffee cherries. The YOLOv8 algorithm was utilized to build a detection model, which was then integrated into a web-based application developed using Streamlit framework. Python was used to implement YOLOv8 and support real-time object detection. The model development process followed the CRISP-DM approach, while the application development adopted a prototyping method. The dataset consisted of 100 primary images collected from Kebun Raya Bogor and 4547 secondary images from Roboflow, divided into 3253 training images, 930 validation images, and 464 testing images. The model achieved an overall mAP50 accuracy of 82.9%, with class-wise accuracy of 90.2% for dry, 76.2% for ripe, 80.9% for unripe, and 84.3% for half-ripe coffee cherries, exceeding the success criteria of 80%. The developed application provides features for detecting coffee cherry ripeness through image uploads and real-time detection using a camera. Usability testing conducted with 16 respondents using the System Usability Scale (SUS) resulted in an average score of 90, classified as "Excellent" with a grade of A. This indicates that the application is highly usable and effectively supports users in detecting coffee cherry ripeness.]]>
Copyrights © 2025
